Bonfy Blog

Cyan data pathways pass through a transparent hub with a broken evidence trail, while an independent audit trail remains intact in the foreground.
AI Governance , AI Agent Security , Access Control , Third-Party Risk

When the Evidence Provider Becomes the Breach

Gidi Cohen Gidi Cohen
Cyan activity trails vanish into a dark monitoring gap, with data blocks beyond, symbolizing missing evidence of AI agent access.
AI Governance , AI Agent Security , Audit Trails , Third-Party Risk

Chainstack Breach Reveals the AI Evidence Gap

Gidi Cohen Gidi Cohen
Damaged cyan database beside an intact reflection, symbolizing AI agents concealing data loss through misleading incident reports.
Data security , AI Governance , AI Agent Security , Audit Trails

AI Coding Agents Are Deleting Databases and Lying About It

Gidi Cohen Gidi Cohen
Glowing cyan data stream bypassing a closed security gate, symbolizing an AI agent crossing access boundaries to reach protected data.
AI Governance , AI Agent Security , Access Control , Healthcare Data Security

A Sanctioned AI Agent Just Breached Medicare

Gidi Cohen Gidi Cohen
AI agent data streams escaping a test environment through an open port into production servers.
MCP Security , AI Governance , AI Agent Security , Prompt Injection

AI Agents Are Now Breaching Production Systems

Gidi Cohen Gidi Cohen
Distinct AI agent signals pass through an authorization gateway along traceable blue paths, while an unauthorized amber path is blocked.
AI Agent Security , Identity Governance , Authorization , NIST

NIST Is Standardizing AI Agent Identity and Authorization

Gidi Cohen Gidi Cohen
AI agent network with blue data pathways passing through an authorization checkpoint and an amber pathway blocked at the boundary.
Shadow AI , AI Agent Security , Identity Governance , Authorization

AI Agent Identity Sprawl Is an Authorization Crisis

Gidi Cohen Gidi Cohen
AI agent data streams escape a transparent testing sandbox toward external servers, where a layered access gate blocks an unauthorized path.
AI Governance , AI Agent Security , Access Control , Agentic AI

AI Labs Cannot Contain Their Own Test Agents

Gidi Cohen Gidi Cohen
Glowing blue connections from numerous MCP servers converge on a data access gateway, illustrating server sprawl and the need for continuous access control.
MCP Security , AI Agent Security , Identity Security , Data Access Control

MCP Server Sprawl Is Outrunning Agent Security

Gidi Cohen Gidi Cohen
Exposed MCP server connections pass through a data security checkpoint that blocks sensitive data while allowing authorized traffic.
MCP Security , AI Agent Security , Cloud Security , Data Access

Exposed MCP Servers Are Breaching Cloud Infrastructure

Gidi Cohen Gidi Cohen
AI agent inside an intact test chamber connected to external servers through an unrestricted network pathway.
AI Governance , AI Agent Security , Data Access Control , AI Safety

An AI Safety Test Agent Attacked Real Targets Alone

Gidi Cohen Gidi Cohen
A single credential opens a broad data archive, with amber access paths extending beyond a narrow cyan task boundary, illustrating excessive delegated access.
Data Governance , AI Governance , AI Agent Security , Access Control

When the AI Company Itself Cannot Control Data Access

Gidi Cohen Gidi Cohen
Fast cyan and amber data streams pass an identity-review carousel toward a financial data archive, illustrating AI agents outpacing traditional identity security.
Data Governance , AI Governance , AI Agent Security , Identity Security

Offensive AI Agents Just Outran Bank Identity Security

Gidi Cohen Gidi Cohen
A glass prism redirects cyan data into an amber network path, illustrating how hidden instructions can divert an AI agent’s legitimate access.
Data Governance , AI Agent Security , Prompt Injection , Salesforce Agentforce

Nobody Hacked Salesforce. The Agent Just Followed Instructions.

Gidi Cohen Gidi Cohen
An amber thread redirects cyan data from digital records through a small opening to an external network, illustrating hidden prompt injection and data leakage.
Contextual Data Enforcement , Data Governance , AI Agent Security

SalesBleed Proves Configured Access Is Not the Same as Actual Data Risk

Gidi Cohen Gidi Cohen

Let's Chat

© 2026 Bonfy.AI. All rights reserved.

Privacy Policy · Terms of Use